CATEGORICAL Subcommand (LOGISTIC REGRESSION command)

CATEGORICAL identifies independent variables that are nominal or ordinal. Variables that are declared to be categorical are automatically transformed to a set of contrast variables as specified on the CONTRAST subcommand. If a variable that is coded as 0 – 1 is declared as categorical, its coding scheme is given indicator contrasts by default.

  • Independent variables that are not specified on CATEGORICAL are assumed to be at least interval level, except for string variables.
  • Any variable that is specified on CATEGORICAL is ignored if it does not appear either after WITH on the VARIABLES subcommand or on any METHOD subcommand.
  • Variables that are specified on CATEGORICAL are replaced by sets of contrast variables. If the categorical variable has n distinct values, there will be n−1 contrast variables generated. The set of contrast variables associated with a categorical variable is entered or removed from the model as a step.
  • If any one of the variables in an interaction term is specified on CATEGORICAL, the interaction term is replaced by contrast variables.
  • All string variables are categorical. Only the first eight characters of each value of a string variable are used in distinguishing between values. Thus, if two values of a string variable are identical for the first eight characters, the values are treated as though they were the same.

Example

LOGISTIC REGRESSION VARIABLES = PASS WITH GPA, GRE, MAT, CLASS, TEACHER
 /CATEGORICAL = CLASS,TEACHER.
  • The dichotomous dependent variable PASS is regressed on the interval-level independent variables GPA, GRE, and MAT and the categorical variables CLASS and TEACHER.